Public Choice 110: 351 363, 2002. 2002 Kluwer Academic Publishers. Printed in the Netherlands. 351 Expressiveness and voting CASSANDRA COPELAND 1 & DAVID N. LABAND 2 1 Division of Economics and Business Administration, Oglethorpe University, Atlanta, GA 30319, U.S.A.; 2 School of Forestry and Wildlife Sciences, Auburn University, Auburn, AL 36849, U.S.A.; e-mail: labandn@mail.auburn.edu Accepted 8 September 2000 Abstract. It has been suggested that voting may be an expressive action taken without regard to any hope of actually influencing election outcomes on the margin. However, there has been no real-world evidence brought to bear on the question of whether the propensity of an individual to vote and the propensity of that same individual to engage in other forms of expressive behavior are correlated in any statistically meaningful sense. Drawing from longitudinal data found in the National Election Surveys we report compelling evidence of a strong, positive relationship between what we term political expressiveness and the act of voting. It is impossible in the real world to determine whether expressive voting occurs, and if so, the extent to which it operates. Fischer (1996: 172) 1. Introduction One of the longstanding conundrums in the scholarly literature of both political science and economics is the paradox of voting. Theory tells us that rational, self-interested individuals will not vote; the facts about voter turnout (by tens of millions of citizens in state and national elections and by very few citizens in small-number local elections) give lie to the theory. The theory of expressive voting (cf. Brennan and Lomasky, 1993) offers a believable resolution to the paradox. Rational, self-interested individuals sometimes, perhaps often, engage in behavior that is not motivated directly by a benefitcost calculation. Behavior may be expressive, in the sense that the actions taken simply cannot influence an outcome. For example, millions of get-well soon cards are sent each year by individuals who would, no doubt, freely We appreciate helpful comments received from Richard Ault, Richard Beil, Bob Bernstein, Danny Butler, Robert Ekelund, John Jackson, John Sophocleus, and Robert Tollison. We are solely responsible for errors in the analysis.
352 admit that the act of sending such a card will not, indeed cannot, influence the recipient s likelihood or speed of recovery. With respect to voting, the application is straightforward: individuals vote because they are expressing themselves about the candidate(s) and/or issues, not because they expect to alter the outcome of the election. The act of expression has inherent value to the individual. In this paper we explore empirically whether the propensity to engage in expressive behavior is related statistically to voter turnout. Our findings, based on analysis of responses to the 1986 1996 National Election Surveys, reveal that there is a strongly positive correlation between the propensity of an individual to vote and that individual s propensity to engage in certain types of politically expressive behavior. In Section 2 we identify conceptual and practical problems that may plague efforts to examine empirically the link between individuals propensity to engage in expressive behavior and their propensity to vote. We present our models, data and results in Section 3. Concluding comments round out the presentation. 2. Linking expressiveness to voter turnout: Conceptual and practical problems The empirical task that we set for ourselves is, in principle, quite simple: measure the statistical correlation between individuals propensity to engage in expressive behavior and their propensity to vote. The practical difficulty is in identifying reasonable measures of expressiveness. Brennan and Lomasky (1993) suggest one measure the sending of get-well soon cards. There are both conceptual and technical problems with a measure such as this. Conceptually, an individual s propensity to engage in expressive behavior generally may not be an accurate reflection of that same individual s propensity to behave expressively with specific respect to political decision-making. This implies that the researcher who is investigating the empirical relationship between expressiveness and voter turnout is less likely to find such a relationship using a measure of general expressiveness than using a measure that reflects what might be termed political expressiveness. Moreover, it is not clear that the propensity to engage in expressive behavior generally exhibits much variability across political jurisdictions. It seems likely that one would find more variability in expressive behavior defined narrowly, along political dimensions for example, across political jurisdictions than in expressive behavior generally. This raises the possibility that a technical problem will plague empirical work lack of variability in a proposed explanatory variable that proxies expressiveness may mitigate the finding of a statistically significant impact on voter turnout when in fact such a relationship may exist.
353 3. Models, data, results Numerous analyses of the determinants of voter turnout have been conducted, as discussed in the review article by Aldrich (1993). Factors that have been found to influence voter turnout significantly include: expected closeness of the election; costs of voting, including opportunity cost of time; age, educational attainment and other demographic variables. Building on this well-established literature, we formulated our general model of the determinants of voter turnout to include a measure of voter expressiveness and a measure of instrumentalism: Turnout = f(expressiveness, closeness of race, other established factors). (1) The theoretical relationship between expressiveness and voting has been posited to be positive more expressive individuals are expected to be more likely to vote than less expressive individuals. Instrumental voting implies that the propensity for an individual to vote will increase as the likelihood that his/her vote will be decisive increases, ceteris paribus. This likelihood is related inversely to: (1) the expected margin of electoral victory, and (2) the expected number of voters. 1 In addition, the propensity to vote is expected to be influenced in predictable ways by factors such as income, age, education, employment status, etc. factors that influence the opportunity cost of time spent voting. 3.1. The National Election Surveys Our first empirical examination was based on analysis of responses to the National Election Surveys conducted in 1986, 1988, 1990, 1992, 1994, and 1996 by the University of Michigan s Center for Political Studies. 2 Survey respondents were required to be United States citizens and eligible voters as of 5 November of the election year in question. The Surveys are distributed under a three-stage selection process to ensure geographic and Census block representation and include numerous demographic questions in addition to the questions about politics (party identification, feelings about political issues and candidates, voting behavior, etc.). Equation (2) identifies the specific model we estimated:
354 VOTE i = a 0 + a 1 AGE i + a 2 4 INCOME i + a 5 7 SCHOOL i + a 8 MARRIED i + a 9 MALE i + a 10 11 RACE i + a 12 EMPLOYED i + a 13 15 REGION i + a 16 UNION i + a 17 IMPACT i (2) + a 18 FEC GIFT i + a 19 BUTTON i + a 20 NOSAY i + a 21 CLOSE i + a 22 CARE i + ε i, where VOTE i = 1 if respondent i voted, 0 if (s)he did not vote; AGE i = the respondent s age; INCOME i = annual family income, defined in three categories: $20,000 $49,999, $50,000 $74,999, and $75,000 and over. Income less than $20,000 is the omitted control category; SCHOOL i = Educational attainment of respondent i, defined as: completed high school degree, completed baccalaureate degree, and completed post-baccalaureate degree, with the omitted control group being those respondents who did not complete high school; MARRIED i = 1 if respondent i was married, 0 if respondent i was not married (single, divorced, or widowed); MALE i = 1 if respondent i was male, 0 if respondent i was female; RACE i = Two 0-1 dummy variables identifying respondent i as being either white or black, with the omitted control category being non-white/non-black (i.e., indian, asian, pacific islander, etc.); EMPLOYED i = 1 if respondent i was employed full-time, 0 if respondent i was not employed full-time; REGION i = Three 0-1 regional dummy variables for the Northeast, the North-Central, and South regions, with the West being the omitted control region. UNION i = 1 if respondent i reported being a union member, 0 otherwise; FEC GIFT i = 1 if respondent i answered yes to the question, Did you make a $1 donation to the Federal Election Commission on your individual income tax returns?, 0 if respondent answered no to this question;
BUTTON i = 1 if respondent i answered yes to the question, Did you wear a sticker, button, or place a sign in your yard during the months before the election?, 0 if (s)he answered no to this question; NO SAY i = 1 if the respondent agreed strongly or somewhat with the statement, People like me don t have any say about what the government does; 0 if the respondent disagreed or was indifferent to this statement. CLOSE i = 1 if the respondent indicated that the presidential race will be close ; 0 if (s)he indicated that one of the candidates will win by quite a bit. CARE i = Respondents were asked, Generally speaking, would you say you personally care a good deal who wins the presidential election this fall? We coded the variable with a 1 if the respondent cared a good deal ; 0 if (s)he did not care very much who wins. ε i = the disturbance term. 355 The propensity for an individual, i, to vote is expected to be positively related to age, for at least three reasons. First, wealth tends to increase with age, giving more incentive for the individual to vote on matters that might affect his/her wealth. Second, and relatedly, individuals might be more willing to express opinions (through voting) about policy issues that touch their wealth than to express opinions on non-wealth related issues. Finally, the opportunity cost of time spent in information acquisition and voting is lower for retired people than for working people. Educational attainment and union membership have been found by numerous researchers (e.g., Campbell, Converse, Miller, and Stokes, 1976; Conway, 1991; Rosenstone and Hansen, 1993) to be positively related to propensity to vote. Marriage tends to increase one s social contacts and interactions that result in increased propensity to vote by reducing the costs of obtaining political information. The empirical findings with respect to the impact of race and gender on propensity to vote are inconsistent (Wolfringer and Rosenstone, 1980; Rosenstone and Hansen, 1993). The theoretical relationship between both income and employment and propensity to vote is inconclusive. On the one hand, increased income (employment) may expand the opportunities for exposure to political information. On the other hand, increased income (employment) may imply higher opportunity costs of voting. We included the regional dummies to capture the possible influence of the media calling an election before polls on the west coast are closed, based on exit polling in the east. For the presidential election years of 1988, 1992, and 1996, we controlled for the pre-election estimation of the respondent s
356 perceived closeness of the presidential race. These variables serve as proxies for the likelihood that the individual s vote will prove influential in the election in question, and thus motivate turnout (Aldrich, 1993; Bernstein and Packard, 1997; Shachar and Nalebuff, 1999). The National Election Surveys contained two questions that we employed as proxies for political expressiveness: (1) whether the respondent displayed a button/sticker/sign prior to the election (asked in all six Surveys), and (2) whether the respondent donated money to the Federal Election Commission on his/her federal income tax return (asked on the 1986, 1988, 1990, and 1992 surveys). It seems extremely unlikely that displaying a button/sticker/sign, by itself, will influence the outcome of an election. Therefore, we interpret such behavior as reflecting the individual s expression of opinion about the candidate/issue, rather than being instrumental in determining an outcome. Since money donated to the Federal Election Committee cannot be earmarked to specific candidates, it is simply impossible for such donations to be instrumental with respect to determining specific outcomes. 3 We therefore regard such behavior as purely expressive, in a political context. At issue is whether individuals who exhibit either of these two politically expressive behaviors is more likely to vote than otherwise comparable individuals who do not exhibit either of these behaviors. We used logistic regression to estimate the limited dependent variable (voting) model specified in Equation (2). The coefficient estimates and marginal effects reported in Table 1 are as expected and consistent with the findings of previous researchers. The propensity to vote increases with age, with educational attainment, with income (to a point), and with marriage. The estimated impact of employment on the likelihood of voting is consistently positive, but inconsistently significant. We observe no gender-based differential propensity to vote. Up through 1992, we observe that both whites arid blacks were more likely to vote than non-white/non-black citizens. We find persistent evidence up through the mid-term elections of 1994 that individuals living in the South are less likely to vote than citizens living in western states. 4 Indeed in both 1990 and 1994, citizens living in the west were more likely to vote than citizens living in every other region of the United States. We observe a strong (positive) union effect on voting in the 1992 and 1994 elections, but not in the other election years we examined.
357 Table 1. Logistic regression results from the National Election Surveys: 1986 1996 Variable 1986 1988 1990 1992 1994 1996 Constant 4.088*** 3.608*** 4.946*** 4.071*** 3.055*** 3.132*** 1.018 0.631 1.225 0.557 0.724 0.446 Age 0.053*** 0.040*** 0.046*** 0.036*** 0.043*** 0.035*** 0.013 0.007 0.011 0.005 0.010 0.005 High School 1.057*** 0.826*** 0.857*** 1.230*** 1.025*** 1.028*** 0.263 0.145 0.212 0.168 0.243 0.146 College Degree 1.718*** 1.855*** 1.460*** 2.090*** 1.841 1.837*** 0.428 0.325 0.362 0.286 0.436 0.262 Higher Degree 1.926*** 2.241*** 1.592*** 3.030*** 2.213*** 2.450*** 0.479 0.392 0.394 0.415 0.525 0.349 $20,000 < I 0.163 0.486*** 0.145 0.445*** 0.290* 0.417** < $49,999 0.041 0.085 0.036 0.061 0.069 0.059 $50,000 < I 0.487** 0.948*** 0.177 0.753*** 0.605*** 0.948*** < $74,999 0.121 0.166 0.044 0.103 0.143 0.135 $75,000 < I 0.030 0.747 0.317 0.515 0.464* 0.870*** 0.007 0.131 0.079 0.704 0.110 0.124 Employed 0.221* 0.095 0.402*** 0.173 0.027 0.274 0.055 0.017 0.099 0.024 0.006 0.039 Married 0.270** 0.374** 0.391*** 0.219 0.378*** 0.509*** 0.067 0.065 0.097 0.030 0.090 0.072 Male 0.047 0.213 0.074 0.137 0.072 0.153 0.012 0.037 0.018 0.019 0.017 0.022 White 0.463 0.813** 1.568*** 0.967** 0.326 0.189 0.115 0.142 0.389 0.132 0.077 0.027 Black 0.987** 1.108*** 1.762*** 1.419*** 0.215 0.205 0.246 0.194 0.436 0.194 0.051 0.029 Northeast 0.256 0.259 0.627*** 0.062 0.668*** 0.181 0.064 0.045 0.155 0.009 0.158 0.026 North Central 0.088 0.091 0.297* 0.331 0.472*** 0.219 0.022 0.160 0.074 0.045 0.112 0.031 South 0.520*** 0.841 0.526*** 0.529*** 0.570*** 0.012 0.130 0.147 0.130 0.072 0.135 0.002 Union 0.085 0.068 0.181 0.619*** 0.485*** 0.210 0.021 0.012 0.045 0.085 0.115 0.030
358 Table 1. Continued Variable 1986 1988 1990 1992 1994 1996 Close Race # 0.242 # 0.215 # 0.036 0.042 0.029 0.005 No Say # 0.673*** 0.224** 0.551*** 0.453*** 0.766*** in Govt. 0.118 0.056 0.075 0.107 0.109 Care # 1.151*** # 1.413*** # 1.256*** 0.202 0.193 0.179 FEC 0.379*** 0.283* 0.426*** 0.083 # # Donation 0.094 0.049 0.106 0.011 Campaign 1.730*** 1.319*** 1.453*** 1.069*** 1.797*** 0.819*** Button 0.431 0.231 0.360 0.146 0.426 0.117 Log likelihood 1003.025 635.846 955.414 721.809 857.028 582.476 Sample size 1756.0 1394.0 1674.0 1781.0 1545.0 1359.0 Significant at the 0.01 level Significant at the 0.05 level Significant at the 0.10 level # question not asked on this Survey. The other variables in the model are of specific relevance to the rational voter and expressive voter models. We find no evidence that predicted closeness of election exerts a significant impact on the likelihood that an individual votes. 5 There are at least two explanations of this result: (1) predicted closeness of election simply does not matter at all, or (2) predicted closeness of election only matters when the two candidates/ballot items are not viewed by prospective voters as close substitutes. We also find consistent evidence that when individuals believe that they have no say in what the government does, they are less likely to vote than individuals who do not believe they have no say in what the government does. This finding is in accord with the rational voter model. For three of the four survey years that included the Federal Election Commission donation question, we find a positive and significant correlation between donating to the Presidential Election Fund and propensity to vote. This is consistent with the notion that individuals vote as a form of political expression. Indeed, this interpretation is reinforced considerably by our findings of a very sizable, positive estimated impact of displaying a campaign button/sticker/sign prior to the election on the propensity to vote, in every
359 election year analyzed. We estimate that individuals who reported displaying campaign buttons/stickers/signs were approximately 40% more likely to vote in the non-presidential election years than individuals who did not report displaying campaign buttons/stickers/signs. The estimated impact is substantially smaller in presidential election years. 4. Concluding comments In reaction to the first draft of this paper, the referee provided the following comment: If one assumes large numbers, rationality, a modicum of knowledge (i.e., people realize that their vote has a zero effect on the voting outcome), and positive costs of voting, then voting is purely an expressive act. End of story. Anyone who seeks to explain voting patterns in this context must perforce proceed in one of two directions. First, one might attempt to shed light on the conditions under which political expressiveness occurs. That is, assuming that individuals have a tendency to be expressive about certain things (including, but not limited to, candidates for political office and/or policy issues), under what conditions will we actually observe expressive behavior like voting? Such analyses would be focused on relative prices and constraints, with believable, if unremarkable, findings such as the fact that observed political expressiveness declines when the cost of being politically expressive rises (e.g., voter turnout is dampened by wet weather). Second, one might attempt to develop insights about expressiveness per se. Here, it seems to us, is where potentially fruitful and important work beckons. As a start in this regard, we identify a statistically significant relationship between the act of voting and two expressive political actions: earmarking tax payments to the presidential election fund and displaying campaign buttons, stickers and/or signs. Our findings are robust across six large-sample national survey populations in different election years. The estimated relationship between one of our measures of political expressiveness, displaying a campaign button, sticker, or sign prior to the election, and the propensity to vote is remarkably large. We estimate that individuals who display campaign buttons/stickers/signs prior to a non-presidential election are approximately 40% more likely to actually vote than individuals who do not display campaign buttons/stickers/signs. The estimated effect is smaller for presidential elections. But there is much work to be done, with respect to improving our understanding of political expressiveness. The empirical correlation that we
360 observed between political expressiveness and voter turnout for national elections may be found to be more or less strong in different electoral contexts. Further, there is the as-yet unexplored question of just how political the propensity to be expressive need be to be correlated with the propensity to vote. Is voting correlated with a general tendency to be expressive or is it correlated strictly with political expressiveness. How strong are such effects, if observed? Finally, if the propensity to vote is found to be correlated with the tendency to be politically expressive, how tight is the relationship between the politically expressive behavior and the election in question? Is the propensity to vote higher when the politically expressive behavior refers specifically to election candidates or issues (e.g., the individual has planted election signs in his/her yard) than it is when the politically expressive behavior is not tied specifically to the election (e.g., the individual flies an American flag on Independence Day)? The answers to these and other related questions will enhance our understanding of the relationship between the act of voting and the propensity to engage in other forms of expressive behavior. Further investigation of this relationship is likely to be scientifically fruitful because models of voter turnout that do not include measures of expressiveness explain a relatively small amount of the variation in turnout (Matsusaka and Palda, 1999). While it is possible that the decision to vote is, essentially, random, it seems doubtful than randomness provides a satisfactory explanation for the persistent decline in voter turnout in national elections during the last half of the 20th century. Further, it seems unlikely that the cost-benefit calculus applied implicitly to the decision to vote in the rational actor model has changed fundamentally over this period. The noted decline in voter turnout may plausibly be due to an erosion in our (general) sense of civic responsibility. However, subjecting this thesis to a rigorous empirical test might prove difficult, because of: (1) the inherent unreliability of survey questions that ask an individual to rate his/her own sense of civic responsibility, and (2) the absence of any commonly-shared standard of civic responsibility that would permit comparisons to be drawn across individuals. However, expressive behaviors can be observed externally. It seems plausible to suggest that declining voter turnout may reflect a generally increasing happiness among prospective voters with their lot in life, with a consequent decline in their motivation to be politically expressive. 6 Notes 1. See especially, Shachar and Nalebuff (1999). One implication of the fact that the likelihood that a vote cast by an individual will be decisive declines with the number of voters
361 is that the time-trend of voter turnout should decline over time. However, across political jurisdictions with differing population growth rates, the time-trend of voter turnout should decline more rapidly in jurisdictions with higher population growth rates. 2. The 1988 (ICPSR 9196), 1992 (ICPSR 6067), and 1996 (ICPSR 6896) national election surveys include information gathered from pre- and post-election interviews, enhanced with previous panel data. 3. For that matter, the identities of possible recipients of the donated money likely are unknown to the donor at the time (s)he files the return. 4. This result is consistent with the findings of Teixeira (1987), and Wolfinger and Rosenstone (1980). 5. Our findings in this regard are consistent with those of Ashenfelter and Kelley (1975) and Foster (1984). 6. This interpretation is supported by the recent work of Cox and Alm (1997) who argue that the real standard of living for Americans generally, including those nominally designated as poor by virtue of some arbitrary government standard, has been rising rapidly during the past half-century. References Aldrich, J.H. (1993). Rational choice and turn-out. American Journal of Political Science 37: 246 278. Ashenfelter, O. and Kelley, S., Jr. (1975). Determinants of participation in congressional elections. Journal of Law and Economics 18: 695 733. Bernstein, R.A. and Packard, E. (1997). Paths to participation in electoral politics: The importance of mobilization. Mimeo. Paper presented at the American Political Science Association meetings in Washington, DC, August. Brennan, H.G. and Lomasky, L.E. (1993). Democracy and decision. Cambridge: Cambridge University Press. Campbell, A., Converse, P.E., Miller, W.E. and Stokes, D.E. (1960). The American voter. New York: Wiley. Conway, M.M. (1991). Political participation in the United States. Washington, DC: Congressional Quarterly Inc. Cox, W.M. and Alm, R. (1997). Time well spent. Federal Reserve Bank of Dallas, Annual Report. Fischer, A.J. (1996). A further experimental study of expressive voting. Public Choice 88: 171 184. Foster, C.B. (1984). The performance of rational voter models in recent presidential elections. American Political Science Review 78: 678 690. Matsusaka, J.G. and Palda, F. (1999). Voter turnout: How much can we explain? Public Choice 98: 431 446. Rosenstone, S.J. and Hansen, J.M. (1993). Mobilization, participation, and democracy in America. New York: Macmillan. Shachar, R. and Nalebuff, B. (1999). Follow the leader: Theory and evidence on political participation. American Economic Review 89: 525 547. Teixeira, R.A. (1987). Why Americans don t vote: Turnout decline in the United States 1960 1984. New York: Greenwood Press. Wolfinger, R.E. and Rosenstone, S.J. (1980). Who votes? New Haven: Yale University Press.
362 Appendix Table A1. Means (std. errors) for National Election Survey data Variable 1986 1988 1990 1992 1994 1996 Age 42.987 44.337 44.283 44.691 45.390 47.578 (16.862) (16.854) (17.916) (17.058) (17.071) (17.100) High School 0.544 0.537 0.513 0.514 0.519 0.502 (0.498) (0.499) (0.500) (0.500) (0.500) (0.500) College Degree 0.186 0.208 0.194 0.238 0.252 0.273 (0.389) (0.406) (0.396) (0.426) (0.435) (0.446) Higher Degree 0.065 0.070 0.061 0.080 0.082 0.099 (0.246) (0.255) (0.239) (0.271) (0.274) (0.298) $20,000 < I 0.445 0.453 0.309 0.430 0.448 0.397 < $49,999 (0.497) (0.498) (0.462 (0.495) (0.497) (0.489) $50,000 < I 0.098 0.111 0.039 0.152 0.164 0.182 < $74,999 (0.297) (0.314) (0.195) (0.359) (0.371) (0.386) $75,000 < I 0.038 0.047 0.020 0.086 0.102 0.135 (0.190) (0.212) (0.139) (0.280) (0.302) (0.341) Employed 0.645 0.661 0.641 0.645 0.650 0.645 (0.479) (0.473) (0.480) (0.479) (0.477) (0.479) Married 0.563 0.570 0.532 0.554 0.534 0.553 (0.496) (0.495) (0.499) (0.497) (0.499) (0.497) Male 0.453 0.456 0.468 0.478 0.472 0.455 (0.498) (0.498) (0.499) (0.500) (0.499) (0.498) White 0.837 0.840 0.834 0.851 0.867 0.865 (0.369) (0.367) (0.372) (0.357) (0.340) (0.341) Black 0.144 0.126 0.137 0.126 0.112 0.109 (0.351) (0.332) (0.343) (0.332) (0.315) (0.312) Northeast 0.155 0.180 0.192 0.190 0.150 0.153 (0.362) (0.384) (0.394) (0.392) (0.357) (0.360) North Central 0.273 0.269 0.251 0.270 0.271 0.272 (0.446) (0.444) (0.434) (0.444) (0.445) (0.445) South 0.365 0.367 0.366 0.346 0.368 0.369 (0.482) (0.482) (0.482) (0.476) (0.482) (0.483) Union 0.216 0.201 0.174 0.171 0.173 0.183 (0.412) (0.401) (0.380) (0.377) (0.378) (0.387) Close Race # 0.735 # 0.822 # 0.532 (0.441) (0.383) (0.499)
363 Table A1. Continued Variable 1986 1988 1990 1992 1994 1996 No Say # 0.142 0.541 0.348 0.537 0.133 in Govt. (0.349) (0.498) (0.476) (0.499) (0.340) Care # 0.654 # 0.779 # 0.790 (0.476) (0.415) (0.408) FEC 0.330 0.284 0.257 0.274 # # Donation (0.470) (0.451) (0.437) (0.446) Campaign 0.073 0.098 0.066 0.121 0.073 0.104 Button (0.260) (0.297) (0.249) (0.327) (0.260) (0.305) N 1756.0 1394.0 1674.0 1781.0 1545.0 1359.0